Exploring the Potential of Life-History Key Innovation : Brook Breeding In

MEC_1543.fm Page 1453 Tuesday, July 16, 2002 11:31 PM

Molecular Ecology (2002) 11, 1453–1463

BlackwellExploring Science, Ltd the potential of life-history key innovation: brook breeding in the radiation of the Malagasy treefrog genus Boophis

M. VENCES,*¶ F. ANDREONE,† F. GLAW,‡ J. KOSUCH,§ A. MEYER,* H.-C. SCHAEFER§ and M. VEITH§ *Department of Biology (Evolutionary Biology), University of Konstanz, 78457 Konstanz, Germany, †Museo Regionale di Scienze Naturali, Laboratory of Vertebrate Taxonomy and Ecology, Via G. Giolitti, 36, 10123 Torino, Italy, ‡Zoologische Staatssammlung München, Münchhausenstrasse 21, 81247 München, Germany, §Institut für Zoologie, Universität Mainz, Saarstraße 21, 55099 Mainz, Germany

Abstract The treefrog genus Boophis is one of the most species-rich endemic amphibian groups of Madagascar. It consists of species specialized to breeding in brooks (48 species) and ponds (10 species). We reconstructed the phylogeny of Boophis using 16S ribosomal DNA sequences (558 bp) from 27 species. Brook-breeders were monophyletic and probably derived from an ancestral pond-breeding lineage. Pond-breeders were paraphyletic. The disparity in diversification among pond-breeders and brook-breeders was notable among endemic Malagasy frogs, although it was not significant when considering Boophis alone. Sibling species which have different advertisement calls but are virtually indistinguishable by morphology were common among brook-breeders; genetic divergence between these species was high (modal 8% total pairwise divergence). Substitution rates in brook-breeding species were significantly higher than in pond-breeders. Speciation of pond-breeders may be hindered by their usually more synchronous reproduction and a higher vagility which enhances gene flow, while a higher potential of spatial segregation and speciation may exist along brooks.

Keywords: Amphibia, Madagascar, Mantellidae, phylogeny, sibling species

Received 28 November 2001; revision received 18 April 2002; accepted 18 April 2002

Recent theoretical and empirical evidence suggests Introduction that rapid diversification processes may in part take place Adaptive radiations have received particular attention as through sympatric rather than allopatric speciation. This sources of organismal diversity. One major topic has been the process is triggered by ecological specialization (Dieckmann exploration of why some lineages gave rise to higher levels & Doebeli 1999) or sexual selection (Higashi et al. 1999; of diversity than others. As an explanation, the influence of key Gavrilets 2000; Boughman 2001). The latter model highlights innovations has often been invoked (Sanderson & Donoghue the importance of optical and other mate recognition 1996). These have been defined as morphological, physiological, mechanisms for speciation (Barraclough et al. 1995). or behavioural traits that enable a lineage to colonize new Treefrogs of the genus Boophis are one of the most adaptive zones and thereby confer the ability to undergo spectacular radiations within the endemic Malagasy family rapid speciation and diversification (Simpson 1953; Liem Mantellidae (Bossuyt & Milinkovitch 2000; Vences & Glaw 1973). 2001). Taking into account recent descriptions, more than 40 nominal and about 18 undescribed Boophis species are attributed to seven phenetic species groups (Blommers- Correspondence: Miguel Vences. ¶Present address: Institute of Biodiversity and Ecosystem Dynamics, Zoological Museum, Schlösser & Blanc 1991; Glaw & Vences 1994, 2000). Among University of Amsterdam, PO Box 94766, 1090 GT Amsterdam, the anurans, the genus Boophis stands out for its large number Netherlands; Fax: + 49 2219417284; E-mail: [email protected] of cryptic sibling species. By morphology alone these are

MEC_1543.fm Page 1454 Tuesday, July 16, 2002 11:31 PM

1454 M. VENCES ET AL.

virtually indistinguishable, and they are usually diagnosed Phylogenetic analyses only by their advertisement call or by differences in coloration (Blommers-Schlösser 1979; Glaw & Vences 2000). Phylogenetic analyses were carried out using paup*, version Speciation through call differentiation has been suggested 4b8 (Swofford 2001). Prior to phylogenetic reconstruction, for certain insect groups in which females recognize their we explored which substitution model fitted our sequence mates by species-specific advertisement calls (Henry 1994). data the best. We applied a hierarchical likelihood method Similarly, in Boophis the bioacoustically divergent cryptic to test the goodness-of-fit of nested substitution models, species may also be examples for this type of speciation. If using the program modeltest (Posada & Crandall 1998). this were the case, a low genetic differentiation among many The substitution model estimated as best fitting our data Boophis species would be predicted, as was documented in (see Results) was used to obtain maximum likelihood (ML) other instances of incipient speciation (e.g. Meyer et al. 1990; trees using the heuristic search option, and a random addition- Schliewen et al. 1994; Freeland & Boag 1999; Wilson et al. sequence with 10 replicates. Maximum parsimony (MP) 2000). and neighbour-joining (NJ) trees were also calculated. In this study, we sequenced fragments of the 16S ribosomal In the MP analysis we conducted two heuristic searches, RNA (rRNA) gene from 27 Boophis species to investigate coding each gap (also in multiple gap sequences) either as their phylogenetic relationships, ages and rates of speciation fifth state or as missing character, with branch swapping using among brook and pond-breeding species. We used the the tree bisection-reconnection (TBR) routine. To identify mitochondrial data to test the hypotheses of (i) low genetic multiple islands of equally most parsimonious trees, 100 divergence between sibling species and (ii) reproductive random addition-sequence replicates were performed. Only mode as possible key innovation and mechanism for minimal-length trees were saved and zero-length branches diversification. were collapsed. In the NJ analyses we used LogDet distances, which are robust against possible variation of sequence evolution among lineages (Lockhart et al. 1994). All MP and Material and methods NJ analyses were repeated after exclusion of the hypervariable region (53 bp). DNA sequencing All available Boophis sequences were included in a first DNA was extracted using QIAmp tissue extraction kits data exploration. To avoid the masking of cladogenetic (Qiagen) from muscle tissue samples preserved in pure patterns in the tests of tree shape and in diversity compar- ethanol. We used the primers 16SA-L (light chain; 5′-CGC isons by inclusion of multiple sequences of the same species, CTG TTT ATC AAA AAC AT-3′) and 16SB-H (heavy chain; we conducted further analyses with a reduced data set of 5′-CCG GTC TGA ACT CAG ATC ACG T-3′) of Palumbi only a single sequence per species. The excluded sequences et al. (1991) to amplify a section [ca. 550 base pairs (bp)] differed by a maximum of only three substitutions from the of the mitochondrial 16S ribosomal RNA gene. We chose included sequences. this gene instead of a more variable marker, such as Two thousand bootstrap replicates (Felsenstein 1985) were cytochrome b or the control region, based on preliminary run in all analyses except ML (only 100 bootstrap replicates data which indicated a high amount of genetic divergence due to computational constraints). The robustness of nodes between Boophis species. Polymerase chain reaction (PCR) was tested by Shimodaira–Hasegawa tests as implemented conditions followed Vences et al. (2000b). PCR products in paup*. Sequences were checked for clock-like behaviour were purified using QIAquick purification kits (Qiagen) using the program tree-puzzle 5.0 (Schmidt et al. 2000). and were sequenced using an automatic DNA sequencer Tests of relative molecular substitution rates (Takezaki (ABI 377). Sequences (see Appendix I for GenBank accession et al. 1995) were conducted using phyltest (Kumar 1996); numbers) were aligned using the computer program of the different substitution models implemented in this sequence navigator 2.0 (Applied Biosystems), taking program, we used Jukes–Cantor distances, uncorrected secondary structures into account (Kjer 1995). The alignment P-distances of transversion only, Kimura two-parameter needed single gaps at four positions, double gaps at two distances, and Kimura two-parameter distances of trans- positions, and three or four consecutive gaps at two versions only. additional positions. Of the latter two cases, one was due to insertions in only four species, while the second was within Tests of diversification rate and tree imbalance a hypervariable region of 53 bp (positions 257–309 of the alignment). Although disparities in species numbers among lineages In addition to our own data we also re-analysed the five may seem obvious, the differences are not significant in many Boophis sequences from the work of Richards et al. (2000) cases. Under a Markovian model of cladogenesis a stochastic who studied higher-level relationships in a more inclusive difference in initial rates can lead to a much higher diversity sample of Malagasy frogs. in one of the lineages (Slowinski & Guyer 1989). Careful

MEC_1543.fm Page 1455 Tuesday, July 16, 2002 11:31 PM

RADIATION OF TREEFROGS IN MADAGASCAR 1455

examination of the initial data is therefore needed to assess (Sanderson 1997). powell algorithm was chosen and the unit possible differences in species diversity (Sanderson & for branch lengths was defined as per site. Donoghue 1996). Our approach largely follows Cook & Lessa (1998) who applied a number of different statistics to Results a similar problem (the assumed high rates of diversification in subterranean rodents). Standard null model tests Phylogeny and substitution rates (Slowinski & Guyer 1989, 1993) were used to compare standing diversity in brook-breeding and pond-breeding In the preliminary data exploration using all 37 available Boophis. The more sensitive three-taxon tests, as implemented Boophis sequences and three outgroups, and coding gaps as in lrdiverse (Sanderson & Donoghue 1994; Sanderson & fifth state, of 558 characters included in the analysis, 328 Wojciechowski 1996), were performed with 1000 Monte were constant, 44 were variable but parsimony-uninformative, Carlo replicates, keeping the relative age of the internal and 186 were parsimony-informative. modeltest suggested node as unknown. a general time-reversible substitution model with a gamma We used the Colless index (Colless 1982) and confidence distribution shape parameter of 0.713, a proportion of intervals as provided by Kirkpatrick & Slatkin (1993) as a invariable sites of 0.498 and empirical base frequencies measure of tree imbalance. This index was calculated by hand (A: 0.341; C: 0.207; G: 0.196; T: 0.256) and substitution rates from the topology of the ML tree. The program End-Epi (A-C, 3.65; A-G, 17.030; A-T, 7.938; C-G, 2.380; C-T, 39.395; (Rambaut et al. 1997) was used to estimate relative clado- G-T, 1.000). These settings were used for ML analyses. genesis. This statistic calculates the probability of the exist- In the ML tree (Fig. 1), with the exception of the Boophis ence of a given number of tips (terminal taxa) of a lineage majori group, pairs and trios of sibling species were through time in comparison with the actual number of tips, grouped as monophyletic units. Species assigned to the and identifies branches with higher than expected rates of same phenetic species group formed monophyletic groups cladogenesis (Nee et al. 1996). As a basis for calculation of in some cases (B. goudoti group; B. luteus group) but appeared relative cladogenesis (see below) we built a tree using the to be paraphyletic in several other instances (B. majori group; nonparametric rate-smoothing algorithm in the r8s program B. rappiodes group; B. tephraeomystax group). The included

Fig. 1 Maximum likelihood phylogram based on all available sequences of the 16S rRNA gene (558 bp). Rana temporaria was used as outgroup. Numbers are bootstrap values in per cent (2000 replications) for MP and NJ analyses, respectively. No values are given for nodes which received support below 50% in both analyses. Values above 70% are printed in bold. The arrow marks the support for monophyly of the brook-breeding clade. Different symbols mark pairs or trios of sibling species which are not or are almost not distinguishable morphologically after preservation (Table 1). Sequences of conspecific specimens are included in boxes.

MEC_1543.fm Page 1456 Tuesday, July 16, 2002 11:31 PM

1456 M. VENCES ET AL.

Table 1 Classification of pairs and trios of sibling species (not or almost not distinguishable morphologically after preservation) included in the present study relative to their distribution pattern and bioacoustic differences

Altitudinally separated (Potentially) syntopic Allopatric species with species with weak call species with strong Allopatric species with weak call differences differences call differences strong call differences

B. luteus B. ankaratra B. marojezensis B. jaegeri — B. septentrionalis — B. aff. ankaratra — B. majori — B. luteus — B. aff. majori B. tephraeomystax B. erythrodactylus B. jaegeri — B. doulioti — B. aff. erythrodactylus — B. septentrionalis B. rappiodes — B. aff. rappiodes B. mandraka — B. aff. mandraka 1* — B. aff. mandraka 2*

*B. mandraka 1 and 2 refer to two undescribed species which both are morphologically similar to B. mandraka.

brook-breeding Boophis were grouped in one clade, whereas support brook-breeder monophyly, but in a 50% majority the pond-breeding species were arranged paraphyletically consensus tree brook-breeding Boophis were monophyletic at the basis of the cladogram. Sequences from specimens (79%). The same was true in a bootstrap analysis (59%). which a priori were considered to be conspecific were Exclusion of the hypervariable region resulted in 200 equally grouped with high bootstrap support as monophyletic most parsimonious trees (657 steps, CI 0.385, RI 0.627). In a units in all cases except for B. tephraeomystax. strict consensus of these the brook-breeding species did These results were also confirmed by the different NJ form a monophyletic group, which received a bootstrap and MP analyses performed. value of 79%. Monophyly/paraphyly of Boophis species (i) NJ, both with all characters and after exclusion of groups and relationships of sibling species were as in the the hypervariable region, supported the same patterns former analyses. outlined above; bootstrap values for the analysis including In the reduced set of 27 Boophis sequences (one sequence all characters are given in Fig. 1, those after exclusion of the per species), coding gaps as fifth character and not including hypervariable sites did not show relevant differences. the hypervariable region, 328 out of 558 characters were (ii) MP, with gaps coded as fifth state and all characters constant, 54 were variable but parsimony-uninformative, included, yielded 12 equally most parsimonious trees and 176 were parsimony-informative. The topology of the [1042 steps, consistency index (CI) 0.378, retention index ML tree as shown in Fig. 2 totally agreed with the results (RI) 0.610]; these were not unequivocal regarding brook- of the preliminary data exploration. As in the previous breeder monophyly, since a second alternative, placing analyses, all MP and NJ searches (with hypervariable region basally a clade containing Boophis rappiodes, B. viridis and B. included or excluded) agreed upon the monophyly/para- erythrodactylus was also equally parsimonious. However, phyly of the different Boophis species groups. Monophyly the bootstrap value (60%) and a majority rule consensus of brook-breeding species was supported by the NJ trees, tree (67%) did support monophyly of brook-breeding MP and NJ bootstrap analyses (68–69%), and MP 50% Boophis. Excluding the hypervariable region did not alter majority rule consensus trees, while the MP strict consensus these results; 333 equally most parsimonious trees were trees were equivocal in this respect. obtained (684 steps; CI 0.393, RI 0.627). Brook-breeder The sequences of Richards et al. (2000) included four monophyly was not apparent from a strict consensus tree species of brook-breeding Boophis (B. albilabris, B. erythro- but from a majority rule tree (85%) and from the bootstrap dactylus, B. luteus, B. madagascariensis) and one pond-breeder analysis (65%). The monophyly of the included species of (B. doulioti, as B. tephraeomystax). Re-analysis of these sequences the B. luteus group and B. goudoti group and paraphyly of using Aglyptodactylus as outgroup (results not shown) the B. rappiodes group, B. majori group and B. tephraeomystax yielded bootstrap supports of 74%, 85% and 73% (ML, MP, group, and the sister-group relationships of sibling species, NJ) for the monophyly of a group composed by the four were confirmed by each of the analyses. brook-breeders. (iii) MP, with gaps coded as missing data and including A tree obtained enforcing a molecular clock with tree- all characters, yielded 28 equally most parsimonious trees puzzle was rejected on a significance level of P < 0.05 by a (993 steps; CI 0.363, RI 0.605). These did not unambiguously likelihood ratio test when compared with the ML tree. Tests

MEC_1543.fm Page 1457 Tuesday, July 16, 2002 11:31 PM

RADIATION OF TREEFROGS IN MADAGASCAR 1457

Fig. 2 Maximum likelihood phylogram of analysis of a reduced data set (one sequence per species only). Numbers are bootstrap values in per cent (100 replications). Thicker lines indicate lineages that were found to have statistically higher than expected rates of cladogenesis (P < 0.01) after nonparametric rate smoothing of the tree. The arrow marks the support for monophyly of the brook-breeding clade.

of relative nucleotide substitution rates using phyltest conservative approach assuming their monophyly was there- significantly rejected rate constancy when comparing the fore also used. This lrdiverse analysis did not indicate brook-breeding lineage separately with each pond-breeder. significant differences in speciation rate between pond- This result was obtained independently from the substitution breeding and brook-breeding Boophis when the respect- model used, except for a single comparison considering tran- ive standing diversities were compared and Aglyptodactylus sversions only. In all cases the branch lengths of brook- (with three species) was used as outgroup (P-value for null breeders were longer, indicating a faster substitution rate. model = 0.863). Such a conservative assumption was also The results were similar when comparing the pond-breeding used for tests of standing diversity (Table 2). These did not B. doulioti with the four brook-breeders of the data set of Richards provide evidence for an increased diversification of brook- et al. (2000), but in these analyses the comparisons based on breeding Boophis. Cumulative tests also failed to indicate such transversions only did not yield significant differences. a pattern when the second large group of Malagasy frogs with brook specialists, the mantellines, were included in a combined analysis (Table 2). Diversification rates The Colless index of the ML tree (0.391) was above Three taxon tests with the program lrdiverse, using the the high cut-off for the corresponding confidence interval ML tree topology (Fig. 2), indicated a significantly higher (Kirkpatrick & Slatkin 1993) and thus demonstrated a speciation rate in brook-breeders. For instance, using the significant tree asymmetry. pond-breeders closest to the brook-breeding lineage (Boophis Analysis of the clock-like ML tree with End-Epi (Rambaut opisthodon as outgroup, B. pauliani as first ingroup and the et al. 1997) showed significantly higher than expected rates brook-breeders as second ingroup — species diversities 1, 1 of cladogenesis at almost all basal positions of the tree, and 22), only the model assuming different speciation rates starting with the split between the outgroup and Boophis, of the brook-breeding lineage received P-values smaller and including the origin of the main brook-breeding lineages than 0.98 and was therefore selected (P < 0.005). (not shown). The same pattern was found after application However, the nonmonophyly of pond-breeders was only of nonparametric rate smoothing using the r8s program weakly indicated by the phylogenetic results, and a more (Fig. 2).

1458 M. VENCES ET AL.

Table 2 Species diversity in Malagasy anurans and P-values for two-taxon tests of imbalance (Slowinski & Guyer 1989, 1993) and the cumulative test (Slowinski & Guyer 1993) across pairs of taxa

Pond-breeding Brook-breeding P-value

Mantellidae Boophinae (Boophis) 10 48 0.175 Laliostominae (Aglyptodactylus + Laliostoma)4 0 — Mantellinae A (Mantidactylus + Mantella)*† 20 38 0.351 Mantellinae B (Mantidactylus + Mantella)*‡ 27 50 0.355 Hyperoliidae Hyperoliinae (Heterixalus)100— Microhylidae§ Scaphiophryninae 9 0 — Dyscophinae 3 0 — Cumulative tests Boophinae + Mantellinae A 0.234 Boophinae + Mantellinae B 0.235

The phylogenetic basis of the groupings is mainly based on Richards et al. (2000). Following the NJ tree presented by these authors, the subgenus Spinomantis (represented in the phylogeny by Mantidactylus aglavei) is here considered as part of the brook-breeding lineage, and brook-breeding mantellines are considered as a monophyletic group. The species numbers are recent estimates (e.g. Glaw & Vences 2000) that include nominal species and already identified but not yet formally described species (F. Glaw and M. Vences, personal observation). *A few deviating species (brook-breeders in the pond-breeding lineage) were excluded: Mantella cowani group, Mantidactylus grandisonae. †Tree-hole or phytotelmic-breeders in the pond-lineage excluded; all species from the brook-lineage which probably have no free- swimming tadpoles excluded. ‡Tree-hole and phytotelmic-breeders included in the pond lineage; species without free-swimming tadpoles but which regularly reproduce along brooks included in the brook-lineage. §Microhylids of the subfamily Cophylinae are not regarded here; this endemic and speciose radiation (about 50 identified species) has nonfeeding tadpoles and reproduces in terrestrial nests or tree-holes.

An overall comparison of the native frog fauna in (e.g. Avise et al. 1998), and pairwise ML distances, which Madagascar (Table 2) showed that only two out of six sub- better account for the specific substitution model in our families include brook-breeding lineages. For one of these, data set. We calculated divergences (i) between Boophis and the Boophinae (genus Boophis), our data indicated mono- other mantellid genera (Mantidactylus and Aglyptodactylus), phyly of brook-breeders as opposed to pond-breeders. The (ii) between pond- and brook-breeding Boophis, (iii) between second group including both reproductive strategies, the brook-breeding species groups, and (iv) within species subfamily Mantellinae (sensu Vences & Glaw 2001), also groups. In the ML distance calculations the mean intergeneric appears to be divided into two main lineages that largely values were distinctly higher than the intrageneric Boophis correspond to brook and pond breeders (Richards et al. 2000; values (0.42 vs. 0.28, 0.22 and 0.17 ). This trend was less distinct Vences et al. 2002a). If vicariance scenarios apply to explain in the total divergence calculations (16% vs. 13%, 12% and the origin of Malagasy anurans (Bossuyt & Milinkovitch 10%; Fig. 3), probably indicating the start of saturation. Total 2001), each anuran group was present on Madagascar since divergences below 9% were always found between spe- its separation from Gondwana and should have undergone cies a priori assigned to the same phenetic species group. a similar biogeographic history. A direct comparison of Most of these (and all below 6%) referred to species pairs brook-breeding and pond-breeding lineages may therefore considered as sibling species. Divergences between sibling be justified and supports a significantly higher diversity of species ranged from 2 to 13%, with a modal value of 8%. pond-breeders as compared with brook-breeding lineages Species with strong differences in advertisement calls (Mann–Whitney U-test; P < 0.05). (whether syntopic, potentially syntopic, or allopatric) had divergences of 5–13%. The two identified divergences lower than 5% referred to allopatrically separated Levels of divergence species pairs with weak call differences (B. aff. ankaratra — We assessed levels of divergence using two models: pairwise B. ankaratra; B. tephraeomystax — B. doulioti ). Conspecific total sequence divergence in per cent (Fig. 3), which allows specimens consistently showed a very low degree of sequence for a direct comparison with values in other animal groups differentiation (0–3 substitutions; 0–0.5% divergence).

RADIATION OF TREEFROGS IN MADAGASCAR 1459

groups lack the anterolateral hyoid process which is present in the pond-breeding B. tephraeomystax and B. doulioti; and (ii) all brook-breeders are characterized by the apomorphic state of two tarsal elements while there are three at least in one pond-breeder (B. doulioti) (Vences et al. 2002a). Our sample of species (five pond-breeders vs. 22 brook- breeders; 19% vs. 81%; including taxa from all phenetic species groups) was representative of the overall species diversity so far identified in Boophis (Table 2; 17% vs. 83%). As our sample included representatives of all species groups we suppose that our phylogenetic conclusions can also be extended to the remaining species. Four pond-breeders (B. opisthodon, B. doulioti, B. tephraeomystax, B. xerophilus) are known to occur at least partly outside densely forested areas; at least the latter three occur outside rainforest. In the brook-breeding lineage, only B. goudoti lives outside forested areas in the Malagasy high- lands (Blommers-Schlösser 1979) and B. laurenti, B. microtympanum and B. williamsi are specialized to montane heathlands (Glaw & Vences 1994). A few other species, such as B. luteus, are also regularly found in largely degraded areas. The percentage of forest species (which depend at least on stretches of continuous gallery forest) is thus below 70% among pond-breeders and about 90% among Fig. 3 Distribution of pairwise total divergences in per cent among brook-breeders, indicating a possible association of brook- 558 bp fragments of the 16S rRNA gene: (a) between different types breeding with rainforest habitat. of Boophis sibling species (as listed in Table 2): white bars, latitudinally separated species with strong call differentiation; black bars, latitudinally separated species with weak call differentiation; High levels of divergence between siblings dark grey bars, altitudinally differentiated species with weak call We herein provide the first survey of genetic divergence in differentiation; light grey bars, syntopic species with strong call differentiation; (b) between brook-breeding Boophis included in a representative sample of pairs of sibling species in Malagasy the same species group (the sibling pairs are a subset of these frogs. The high genetic divergence observed between most data); (c) between brook-breeding Boophis included in different species pairs contrasts with the low intraspecific variation. species groups; (d) between species of the brook-breeding lineage In six species in which more than one specimen was and the pond-breeding assemblage in Boophis; and (e) between examined (B. ankaratra, B. occidentalis, B. rappiodes, B. cf. sibilans, species of Boophis and outgroup species of Mantidactylus and B. tephraeomystax, B. doulioti), divergence between sequences Aglyptodactylus. Based on the reduced data set (one sequence per did not amount to more than 0.5% (three substitutions). species). The modal divergence of 8% between sibling species was much higher than would be expected in the case of recent Pleistocene–Holocene speciation events (Klicka & Zink Discussion 1997; Avise et al. 1998). We did not identify a single example of a very recently diverged pair of sibling species Monophyly and diversity of brook-breeding Boophis which differ by advertisement call. Instead, all syntopic The present study provides moderate support (bootstrap species with important bioacoustic differentiation had values 60–83%) for the monophyly of the assemblage of high sequence divergences, and the lowest levels of brook-breeding frogs of the genus Boophis. Similar values divergence were found between allopatric species. Our were found by re-analysis of the sequences from Richards data therefore do not provide evidence for a recent rapid et al. (2000) in a reduced taxon set (73–85%). Although sympatric speciation by sexual selection in Boophis. If the Shimodaira-Hasegawa (SH) tests could not exclude signi- same factors influencing Boophis speciation have been ficantly all alternative topologies (results not shown), we active over the evolutionary history of the genus, our data consider the monophyly of brook-breeders as relatively would rather favour allopatric speciation as a prevailing well established since it agrees with two relevant osteological mechanism in this group, as it is assumed for tropical characters: (i) representatives of all brook-breeding species salamanders (García-Paris et al. 2000).

1460 M. VENCES ET AL.

permanent brooks in Madagascar is closely related to Patterns of diversification in Boophis rainforests. Furthermore, most Malagasy rainforests are on Using the molecular phylogeny of Boophis as the basis, the relatively steep slopes where few lentic waters occur. It is three taxon test supports a significant disparity in diversification plausible to assume that rainforests were a novelty for the rates between brook- and pond-breeding Boophis. Also, the ancestors of the extant Malagasy amphibian fauna: In ML tree (Fig. 2) was found to be significantly asymmetric. vicariance scenarios (Duellman & Trueb 1986; Bossuyt & In contrast, other analyses led to less unequivocal results. Milinkovitch 2001) ancestral mantellids were present on Higher than expected rates of cladogenesis were found Madagascar already in the Early Cretaceous when probably in the initial radiation of brook-breeding Boophis, but also a rather seasonal and partly xeric climate prevailed (Spicer characterized all other basal splits in the genus, including et al. 1994). In dispersal scenarios (Krause et al. 1997) they pond-breeding species. This was true when enforcing a reached the Malagasy west coast in the Cenozoic from clock-like tree, but also after nonparametric rate smoothing Africa via waif dispersal, and had to be adapted to aridity which has been shown to provide more accurate estimates to survive under the conditions encountered (Vences et al. of divergence times in the case of nonclock-like substitution 2000a). Hence, the wealth of new ecological niches in rates (Sanderson 1997). rainforest opened up for these ancestral frogs, and only Despite the obvious differences between species numbers some of them succeeded in colonizing this new habitat. of brook-breeding and pond-breeding Boophis species, the Those which switched to reproduction in brooks rather than comparisons of standing diversities were not significant ponds were relieved of the need of synchronous breeding. in a conservative approach that assumed monophyly of Temporal and spatial segregation of subpopulations reduced pond-breeders as an alternative hypothesis. The crucial point effective population sizes and favoured fixation of muta- is the supposed paraphyly of pond-breeders, a question which tions, thereby accelerating DNA substitution rates and cannot be unequivocally solved with the present data. probably promoting speciation. Future analyses of larger data sets are needed to understand the relevance of the indications of a higher diversification Acknowledgements rate of brook-breeding Boophis compared to pond-breeders. We are grateful to Nirhy Rabibisoa, Domoina Rakotomalala, Olivier The tendencies of an overall higher species diversity of Ramilison, Noromalala Raminosoa, Fara Ranaivojaona, Jasmin brook-breeding clades among Malagasy frogs, however, Randrianirina and Denis Vallan for their help in the field. Michael appear to correlate with a lower mitochondrial nucleotide J. Sanderson provided crucial software. We are also indebted to the substitution rate in two pond-breeding lineages: Boophis and Malagasy authorities for research and export permits. The work of the microhylid genus Scaphiophryne (Vences et al. 2002b). F.G. and M.V. was made possible by a cooperation accord between Reproduction in ponds is usual in open savannah landscapes the Département de Biologie Animale, Université d’Antananarivo in which permanent brooks are less frequent. Lentic waters and the Zoologische Staatssammlung München; that of F.A. by a cooperation accord between the Parc Botanique et Zoologique de are often temporary, and a highly synchronous ‘explosive’ Tsimbazaza, Antananarivo, and the Museo Regionale di Scienze breeding behaviour is therefore shown by pond-breeding Naturali, Torino. Financial support was granted by the Deutscher Boophis in arid environments (Andreone et al. in press). Fur- Akademischer Austauschdienst (DAAD) and the Deutsche For- thermore, savannah-breeders are also vagile since ponds schungsgemeinschaft (BO 682/5–1 and VE 247/1–2). may form at different sites each year. Such natural history patterns contribute to a high degree of gene flow within and between populations. This may prevent sympatric and References allopatric speciation and fixation of mutations, thereby Andreone F, Vences M, Guarino FM, Glaw F, Randrianirina J (2002) causing a lower species diversity and slower substitution Natural history and larval morphology of Boophis occidentalis rates in pond-breeders. (Anura: Mantellidae: Boophinae) provide new insights into the Slowinski & Guyer (1993) explicitly distinguish between phylogeny and adaptive radiation of endemic Malagasy frogs. Journal of Zoology, London, in press. intrinsic causes of diversity (i.e. morphological novelties Avise JC, Walker D, Johns GC (1998) Speciation duration and that can be seen as key innovations) and extrinsic causes Pleistocene effects on vertebrate phylogeography. Proceedings of (i.e. environmental changes or tectonic activity). If brook- the Royal Society of London B, 265, 1707–1712. breeding adaptations were partly responsible for Boophis Barraclough TG, Harvey PH, Nee S (1995) Sexual selection and diversity, they certainly could be defined as key innovations. taxonomic diversity in passerine birds. Proceedings of the Royal The fact that four out of six lineages of Malagasy frogs did Society of London B, 259, 211–215. not colonize brooks indicates that important modifications Blommers-Schlösser RMA (1979) Biosystematics of the Malagasy frogs. II. The genus Boophis (Rhacophoridae). Bijdragen Tot de of mating and egg-laying behaviour as well as embryonal Dierkunde, 49, 261–312. and larval development are necessary for an original pond- Blommers-Schlösser RMA, Blanc CP (1991) Amphibiens (première breeder to reproduce successfully in brooks. However, brook- partie). Faune de Madagascar, 75, 1–379. breeding has also an extrinsic aspect. Occurrence of small Bossuyt F, Milinkovitch MC (2000) Convergent adaptive radiations

RADIATION OF TREEFROGS IN MADAGASCAR 1461

in Madagascan and Asian ranid frogs reveal covariation between Meyer A, Kocher TD, Basasibwaki P, Wilson AC (1990) Mono- larval and adult traits. Proceedings of the National Academy of phyletic origin of Lake Victoria cichlid fishes suggested by Sciences of the USA, 97, 6585–6590. mitochondrial DNA sequences. Nature, 347, 550–553. Bossuyt F, Milinkovitch MC (2001) Amphibians as indicators of Nee S, Barraclough TG, Harvey PH (1996) Temporal changes in Early Tertiary ‘Out-of-India’ dispersal of vertebrates. Science, 292, biodiversity: detecting patterns and identifying causes. In: 93–95. Biodiversity: a Biology of Numbers and Difference (ed. Gaston K), Boughman JW (2001) Divergent sexual selection enhances repro- pp. 230–252. Blackwell Science, London. ductive isolation in sticklebacks. Nature, 411, 944–948. Palumbi SR, Martin A, Romano S, McMillan WO, Stice L, Colless DH (1982) Phylogenetics: the theory and practice of Grabowski G (1991) The Simple Fool’s Guide to PCR, Version 2.0. phylogenetic systematics II [book review]. Systematic Zoology, Privately published document compiled by S. Palumbi, Dept. 31, 100–104. Zoology, University of Hawaii, Honolulu. Cook JA, Lessa EP (1998) Are rates of diversification in subterra- Posada D, Crandall KA (1998) Modeltest: testing the model of DNA nean South American tuco-tucos (genus Ctenomys, Rodentia: substitution. Bioinformatics, 14, 817–818. Octodontidae) unusually high? Evolution, 52, 1521–1527. Rambaut A, Harvey PH, Nee S (1997) End-Epi: an application Dieckmann U, Doebeli M (1999) On the origin of species by for reconstructing phylogenetic and population processes sympatric speciation. Nature, 400, 354–357. from molecular sequences. Computational Applied Bioscience, 13, Duellman WE, Trueb L (1986) Biology of Amphibians. McGraw-Hill, 303–306. New York. Richards CM, Nussbaum RA, Raxworthy CJ (2000) Phylogenetic Felsenstein J (1985) Confidence limits on phylogenies: an approach relationships within the Madagascan boophids and mantellids using the bootstrap. Evolution, 39, 783–791. as elucidated by mitochondrial ribosomal genes. African Journal Freeland JR, Boag PT (1999) The mitochondrial and nuclear genetic of Herpetology, 49, 23–32. homogeneity of the phenotypically diverse Darwin’s ground Sanderson MJ (1997) A nonparametric approach to estimating finches. Evolution, 53, 1553–1563. divergence times in the absence of rate constancy. Molecular García-Paris M, Good DA, Parra-Olea G, Wake DB (2000) Bio- Biology and Evolution, 14, 1218–1231. diversity of Costa Rican salamanders: implications of high levels Sanderson MJ, Donoghue MJ (1994) Shifts in diversification rate of genetic differentiation and phylogeographic structure for with the origin of angiosperms. Science, 264, 1590–1593. species formation. Proceedings of the National Academy of Sciences Sanderson MJ, Donoghue MJ (1996) Reconstructing shifts in diver- of the USA, 97, 1640–1647. sification rates on phylogenetic trees. Trends in Ecology and Gavrilets S (2000) Rapid evolution of reproductive barriers driven Evolution, 11, 15–20. by sexual conflict. Nature, 403, 886–889. Sanderson MJ, Wojciechowski MF (1996) Diversification rates in a Glaw F, Vences M (1994) A Fieldguide to the Amphibians and Reptiles temperate legume clade: are there ‘so many species’ of Astragalus of Madagascar, 2nd edn, including Mammals and Freshwater Fish. (Fabaceae)? American Journal of Botany, 83, 1488–1502. Vences and Glaw, Köln. Schliewen UK, Tautz D, Pääbo S (1994) Sympatric speciation Glaw F, Vences M (2000) Current counts of species diversity and suggested by monophyly of crater lake cichlids. Nature, 368, endemism of Malagasy amphibians and reptiles. In: Diversité 629–632. et Endémisme de Madagascar (eds Lourenço WR, Goodman SM), Schmidt HA, Strimmer K, Vingron M, von Haeseler A (2000) TREE- pp. 243–248. Mémoires de la Société de Biogéographie, Paris. PUZZLE, Version 5.0. München. Henry CS (1994) Singing and cryptic speciation in insects. Trends Simpson GG (1953) The Major Features of Evolution. Columbia in Ecology and Evolution, 9, 388–392. University Press, New York. Higashi M, Takimoto G, Yamamura N (1999) Sympatric speciation Slowinski JB, Guyer C (1989) Testing the stochasticity of patterns by sexual selection. Nature, 402, 523–626. of organismal diversity: an improved null model. American Kirkpatrick M, Slatkin M (1993) Searching for evolutionary patterns Naturalist, 134, 907–921. in the shape of a phylogenetic tree. Evolution, 47, 1171–1181. Slowinski JB, Guyer C (1993) Testing whether certain traits have Kjer KM (1995) Use of rRNA secondary structure in phylogenetic caused amplified diversification: an improved method based on studies to identify homologous positions: an example of align- a model of random speciation and extinction. American Naturalist, ment and data presentation from the frogs. Molecular Phylogenetics 142, 1019–1024. and Evolution, 4, 314–330. Spicer RA, Rees P, McA, Chapman JL (1994) Cretaceous phyto- Klicka J, Zink RM (1997) The importance of recent Ice Ages in geography and climate signals. In: Palaeoclimates and Their Modelling. speciation: a failed paradigm. Science, 277, 1666–1669. With Special Reference to the Mesozoic Era (eds Allen JRL, Hoskins BJ, Krause DW, Hartman JH, Wells NA (1997) Late Cretaceous Sellwood BW, Spicer RA, Valdes PJ), pp. 69–78. Chapman & vertebrates from Madagascar. Implications for biotic changes in Hall, London. deep time. In: Natural Change and Human Impact in Madagascar Swofford DL (2001) PAUP*. Phylogenetic Analysis Using Parsimony (eds Goodman SM, Patterson BD), pp. 3–43. Smithsonian Insti- (*And Other Methods), Version 4b8. Sinauer Associates, Sunderland, tution Press, Washington and London. MA. Kumar S (1996) PHYLTEST: Phylogeny Hypothesis Testing Software. The Takezaki N, Rzhetsky A, Nei M (1995) Phylogenetic test of the Pennsylvania State University, Pennsylvania. molecular clock and linearized trees. Molecular Biology and Liem KF (1973) Evolutionary strategies and morphological Evolution, 12, 823–833. innovations: cichlid pharyngeal jaws. Systematic Zoology, 22, Vences M, Glaw F (2001) When molecules claim for taxonomic 425–441. changes: new proposals on the classification of Old World Lockhart PJ, Steel MA, Hendy MD, Penny D (1994) Recovering treefrogs. Spixiana, 24, 85–92. evolutionary trees under a more realistic model of sequence Vences M, Glaw F, Kosuch J, Das I, Veith M (2000a) Polyphyly of evolution. Molecular Biology and Evolution, 11, 605–612. Tomopterna (Amphibia: Ranidae) based on sequences of the

1462 M. VENCES ET AL.

mitochondrial 16S and 12S rRNA genes, and ecological bio- speciation in sympatric Nicaraguan crater lake cichlid fishes: geography of Malagasy relict amphibian groups. In: Diversité sexual selection versus ecological diversification. Proceedings of et Endémisme de Madagascar (eds Lourenço WR, Goodman SM), the Royal Society of London B, 267, 2133–2141. pp. 229–242. Mémoires de la Société de Biogéographie, Paris. Vences M, Kosuch J, Lötters S, Widmer A, Jungfer KH, Köhler J, Veith M (2000b) Phylogeny and classification of poison frogs This work is part of a long-term study of the diversity and (Amphibia: Dendrobatidae), based on mitochondrial 16S and biogeography of the amphibians and other vertebrates of Madagascar. 12S ribosomal RNA gene sequences. Molecular Phylogenetics and Miguel Vences is head of the vertebrate section of the Zoological Evolution, 15, 34–40. Museum, University of Amsterdam and assistant professor; until Vences M, Glaw F, Andreone F, Jesu R, Schimmenti G (2002a) April 2002 he was a postdoctoral fellow of the Deutsche Systematic revision of the enigmatic Malagasy broad-headed Forschungsgemeinschaft in the laboratory of Axel Meyer at the frogs (Laurentomantis Dubois, 1980), and their phylogenetic position University of Konstanz, using molecular methods to contribute to in the endemic mantellid radiation of Madagascar. Contributions the understanding of speciation mechanisms in frogs. While his to Zoology, 70, 191–212. taxonomic and ecological work on the Malagasy herpetofauna is largely Vences M, Aprea G, Capriglione T, Andreone F, Odierna G (2002b) carried out in cooperation with Frank Glaw and Franco Andreone, a Ancient tetraploidy and slow molecular evolution in Scaphio- part of the molecular work has so far been performed in the phryne: ecological correlates of speciation mode in Malagasy laboratory of Michael Veith at the University of Mainz, in relict amphibians. Chromosome Research, 10, 127–136. cooperation with Joachim Kosuch and Hans-Christian Schaefer. Wilson AB, Noack-Kunnmann K, Meyer A (2000) Incipient

RADIATION OF TREEFROGS IN MADAGASCAR 1463

Appendix I

Specimens examined and GenBank accession numbers of sequences. All localities are in Madagascar except that of Rana temporaria (Germany)

Species group Species Origin Voucher Genbank accession

— Rana (Rana) temporaria Koblenz ZFMK 69883 AF124135 — Aglyptodactylus laticeps Kirindy ZFMK 64135 AF215329 — Mantidactylus corvus Isalo ZFMK 70494 AF215320 luteus group Boophis ankaratra Mandraka ZSM 400/2000 AJ315909 Boophis ankaratra Manjakatompo ZSM 367/2000 AJ315911 Boophis ankaratra Col des Tapias ZSM 399/2000 AJ315910 Boophis sp. n. (aff. ankaratra) Ranomafana ZFMK 62907 AJ315912 Boophis jaegeri Berara ZSM 413/2000 AJ315914 Boophis luteus Andasibe UADBA 2000.063 AJ315916 Boophis septentrionalis Ambre ZSM 493/2000 AJ315915 Boophis cf. sibilans Ranomena ZFMK 62797 AF215338 Boophis cf. sibilans Vohidrazana ZSM 327/2000 AJ315913 goudoti group Boophis cf. burgeri Marojejy ZFMK 59905 AF215336 Boophis goudoti Col des Tapias not preserved AJ315917 Boophis madagascariensis An’Ala ZFMK 62265 AF215337 microtympanum group Boophis microtympanum Col des Tapias ZSM 393/2000 AJ315918 albilabris group Boophis occidentalis (large) Berara MRSN A2000 AJ314820 Boophis occidentalis (small) Berara MRSN uncat. AJ314819 majori group Boophis majori Vohiparara ZFMK 62672 AF215340 Boophis sp. n. (aff. majori) Andasibe ZSM 313/2000 AJ315922 Boophis marojezensis Vohidrazana ZSM 326/2000 AJ315923 rappiodes group Boophis erythrodactylus Mandraka ZSM 324/2000 AJ314814 Boophis sp. n. (aff. erythrodactylus) Andasibe ZFMK 62888 AF215339 Boophis mandraka Mandraka ZSM 346/2000 AJ315921 Boophis sp. n. (aff. mandraka 1) Vohidrazana ZSM 310/2000 AJ315919 Boophis sp. n. (aff. mandraka 2) Ambavaniasy NMBE 1046008 AJ315920 Boophis rappiodes (male) Andasibe ZSM 347/2000 AJ314815 Boophis rappiode (female) Andasibe UADBA 2000.59 AJ314816 Boophis sp. n. (aff. rappiodes) Andasibe ZSM 344/2000 AJ314817 Boophis viridis Andasibe ZSM 338/2000 AJ314818 tephraeomystax group Boophis doulioti* unknown* USNM 59146 AF026360 Boophis doulioti Isalo ZFMK 70495 AF215332 Boophis doulioti Kirindy ZFMK 66690 AF215334 Boophis opisthodon Cap EstZFMK 70480 AF215331 Boophis pauliani Andasibe ZSM 345/2000 AJ315924 Boophis tephraeomystax Ambanja ZSM 489/2000 AJ312115 Boophis tephraeomystax Nosy Be ZSM 458/2000 AJ312114 Boophis tephraeomystax Cap EstZFMK 66685 AF215333 Boophis tephraeomystax Sambava UADBA 2000.379 AJ312116 Boophis xerophilus Kirindy ZFMK 66705 AF215335

Collection abbreviations are: MRSN, Museo Regionale di Scienze Naturali, Torino; NMBE, Naturhistorisches Museum Bern; UADBA, Université d’Antananarivo, Département de Biologie Animale (numbers are provisional field numbers of F. Glaw and M. Vences); USNM, United States National Museum; ZFMK, Zoologisches Forschungsinstitut und Museum A. Koenig, Bonn; ZSM, Zoologische Staatssammlung, München. *The B. doulioti sequence marked with a star was obtained from GenBank (published by Richards et al. 2000).